Update app.py

app.py

@@ -71,25 +71,18 @@ and output only the code
     )
     return call_visual_llm(prompt)
 
-def replace_visual_placeholders(report_html: str, context:str, initial_query:str, crumbs:str) -> str:
-    """
-    Searches for visual placeholders in the report HTML and replaces each one with the output generated by the visual LLM.
-    Placeholders are expected to have the form:
-    
-    [[Visual Placeholder X: Purpose: <description>]]
-    
-    Returns the modified HTML.
-    """
-    # This regex matches the placeholder and extracts the description after "Purpose:"
-    pattern = r"\[\[Visual Placeholder.*?\]\]"
-    
+def replace_visual_placeholders(report_html: str, context: str, initial_query: str, crumbs: str) -> str:
+    pattern = r"\[\[Visual Placeholder \d+:(.*?)\]\]"  # Regex to match placeholders
+
     def placeholder_replacer(match):
-        placeholder_instructions = match.group(1).strip()
+        placeholder_instructions = match.group(1).strip()  # Extract and strip placeholder instructions
         logging.info(f"Generating visual for placeholder: {placeholder_instructions}")
+        
         # Call the visual generation function:
         visual_html = generate_visual_snippet(placeholder_instructions, context, initial_query, crumbs)
         return visual_html
 
+    # Replace all matches in the HTML
     new_report_html = re.sub(pattern, placeholder_replacer, report_html, flags=re.DOTALL)
     return new_report_html
 
@@ -255,7 +248,8 @@ def perform_serpapi_search(query: str, engine: str, num_results: int = 10) -> li
     results = []
     while attempt < 3:
         try:
-            response = requests.get("https://serpapi.com/search", params=params, headers=get_random_header())
+            headers = {"User-Agent": get_random_header()}
+            response = requests.get("https://serpapi.com/search", params=params, headers=headers)
             if response.status_code == 200:
                 try:
                     data = response.json()
@@ -288,7 +282,8 @@ def perform_serpapi_search(query: str, engine: str, num_results: int = 10) -> li
 
 def process_pdf(url: str) -> str:
     try:
-        r = requests.get(url, headers=get_random_header())
+        headers = {"User-Agent": get_random_header()}
+        response = requests.get("https://serpapi.com/search", params=params, headers=headers)
         r.raise_for_status()
         f = io.BytesIO(r.content)
         reader = PyPDF2.PdfReader(f)
@@ -340,7 +335,11 @@ The report should be very detailed and lengthy — approximately the equivalent
 
 // Special analytical artefacts.
 - Since the generation of visuals (excluding tables) like graph or charts cannot be done through text, create special placeholders in this format:
-[[Visual Placeholder n:...]]
+[[Visual Placeholder n:
+- Purpose of this visual is:...
+- Relevant data to generate it:...
+- Visual guidance:...
+]]
 with n as the reference number
 - in the placeholder, no need to add the references to the source, but make sure ALL of the data pointes required has a source from the learning and reference material hereafter 
 - these placeholders text should contain:
@@ -814,7 +813,8 @@ def iterative_deep_research_gen(initial_query: str, reportstyle: str, breadth: i
                     process_log += f"Extracted PDF content from {url}\n"
                 else:
                     try:
-                        response = requests.get(url, headers=get_random_header())
+                        headers = {"User-Agent": get_random_header()}
+                        response = requests.get("https://serpapi.com/search", params=params, headers=headers)
                         response.raise_for_status()
                         raw_content = response.text
                         process_log += f"Extracted full page content from {url}\n"